>> Date: Tue, 30 Dec 2014 11:46:38 -0500
>> From: Ted Unangst <[email protected]>
>> 
>> RFC 864 specifies the chargen service. The RFC recommends a recognizable
>> pattern, but it was also written at a time when the internet was a
>> kindler, gentler place. A better recommendation for the modern
>> internet would be to produce random data.
>> 
>> Diff follows. IMO the code is easier to reason about without the
>> ring arithmetic. Benchmarking reveals the code is slightly slower now
>> (8MB/s vs 10MB/s), but I think that's an acceptable tradeoff.
>
>Jesus Ted, now you're defenitely beating dead horses.
>
>inetd is disabled by default, within the default inetd.conf chargen is
>disabled.  If people still manage to enable this code, they deserve
>what they get.
>
>Besides I really don't understand what issue this tries to address.
>Are you setting a trap for some poor Linux user who will blindly adopt
>this "OpenBSD made chargen secure" diff on a system with a poor
>arc4random implementation, exposing the internal state of the speudo
>random generator to the internets?
>
>Oh, and Wikipedia calls the existing pattern a "de facto standard".

I also note the existing pattern has peculiar compressibility, which
the random pattern would not.  The idea is dead.

Reply via email to